Overview

Population Characteristics

  • This dataset include 463 confirmed cases admitted from 2018-08-05 to 2019-07-15.

  • 56% women and 2% children ≤5 years .

  • 150 confirmed deaths reported for an average CFR of 32%.

Sampling

Cases Description

Patient Outcomes by Admission Delay

  • Deceased patients (n=147) had a median admission delay of 5 days (IQR 4.5, range 0-39)
  • Recovered patients (n=253) had a median admission delay of 3 days (IQR 4, range 0-32)

Ct kinetics

  • A linear mixed-effects model with random intercepts per participant showed that Ct values significantly increased over time (β = 0.25, p < 0.001), consistent with viral load decline.
  • Participants who died had markedly lower baseline Ct values compared to recovered individuals (β = −8.65, p < 0.001), indicating higher initial viral loads.
  • The interaction between time and outcome was not significant (p = 0.60), suggesting similar Ct kinetics across outcomes.

\[ Ct_{ij} = \beta_0 + \beta_1 (\text{Time from Onset}_{ij}) + \beta_2 (\text{Outcome}_i) + \beta_3 (\text{Time from Onset} \times \text{Outcome}_i) + u_i + \epsilon_{ij} \]

  • (Ct_{ij}): Ct value for participant i at time j
  • (_0): intercept (baseline Ct for recovered)
  • (_1): daily Ct change among recovered
  • (_2): baseline difference (Death vs Recovered)
  • (_3): interaction (difference in Ct trajectory)
  • (u_i): random intercept
  • (_{ij}): residual error

Cases and CFR

Treatments and complications

These complications are inferred from laboratory results :

  • Renal impairment: Creatinine (mg/dL)≥1.1
  • Hyperkaliemia: Potassium (mmol/L)>5.5
  • Hypokaliemia: Potassium (mmol/L)<3.5
  • Rhabdomyolysis: Creatinine kinase (U/L)>1000
  • Severe hypoglycemia: Glucose (mg/dL) <60
  • Severe thrombocytopenia: Platelets (x10⁹/L) <50

Analyses

Approach

flow ImmuneMod Immune modulation DelDirect Deleterious direct effect of virus ImmuneMod->DelDirect +/- ViralLoad Viral load ViralLoad->DelDirect + OrganFail Organ failure DelDirect->OrganFail + Death Death OrganFail->Death + PreExist Pre-existing comorbidity PreExist->OrganFail + Age Age Age->PreExist +/- Sex Sex Sex->PreExist DelayedPres Delayed presentation Sex->DelayedPres Nutrition Nutrition Nutrition->PreExist + DelayedPres->ViralLoad + Vaccination Vaccination Vaccination->ViralLoad - Antiviral Antiviral therapy Antiviral->ViralLoad - Hypoperf Hypoperfusion / shock Hypoperf->OrganFail + Dehydration Dehydration Dehydration->Hypoperf + MetElect Metabolic / electrolyte derangement Dehydration->MetElect + MetElect->OrganFail + Hypoxa Hypoxaemia Hypoxa->OrganFail + Anemia Anaemia Anemia->Hypoxa + SecInf Secondary infection SecInf->OrganFail + SecInf->Hypoxa + Bleeding Bleeding Bleeding->Anemia + ORS ORS ORS->Dehydration - IVfluid IV fluid IVfluid->Dehydration - IVfluid->MetElect - Diarrhoea Diarrhoea Diarrhoea->Dehydration +
  • This Directed Acyclic Graphs (DAG) illustrates hypothesized causal relationships underlying EVD pathophysiology and clinical outcomes.
workflow A 1. Data cleaning and wrangling B 2. Univariate model screening • Identify sparse variables • Exclude predictors causing quasi-complete separation A->B C 3. Collinearity assessment • Variance Inflation Factor (VIF) • GVIF_scaled ≤ 1.68 B->C D 4. LASSO regression • Shrinks weak predictors • Retains strongest associations C->D E 5. Multivariable logistic regression • Include LASSO-selected predictors D->E F OR from multivariable model E->F G Adjusted RR (Poisson) E->G H Raw RR for binary predictors E->H

Ct threshold to predict participant outcome

  • ROC analyses were performed to determine the optimal Ct threshold to predict participant outcome.
  • With an area under the curve (AUC) of 0.85, ROC analyses evaluated the optimal Ct threshold cut-off at 23.25 for predicting patients outcome (sensitivity 79%, specificity 77%).

Independent Predictors of EVD Outcome and Complications

  • Multivariable Poisson regression with robust standard errors was used to estimate adjusted risk ratios (aRR) and 95% confidence intervals.
Variable Relative Risk (95% CI) Direction
CT at Admission 0.86 (0.84-0.89) ⬇
Admission Delay 1.02 (1-1.04) ⬆
Hyperkaliemia 1.27 (0.82-1.97) ↔︎
Rhabdomyolysis 1.16 (0.8-1.67) ↔︎
Anuria 1.36 (0.8-2.32) ↔︎
Difficulty breathing 1.58 (1.2-2.08) ⬆
Signs of dehydration 0.65 (0.47-0.91) ⬇
Unusual Bleeding/Bruising 1.32 (1.07-1.64) ⬆
Anorexia 0.88 (0.65-1.19) ↔︎
Diarrhoea 1.14 (0.85-1.52) ↔︎
Fatigue/Weakness 1.51 (0.89-2.58) ↔︎
Nausea/Vomiting 0.85 (0.6-1.19) ↔︎
Sore throat / Swallowing 1.06 (0.82-1.37) ↔︎
  • Patients with unusual bleeding/bruising had an increased risk of death by 32% (RR 1.32, 95% CI: 1.7-1.64).
Variable Relative Risk (95% CI) Direction
CT at Admission 0.97 (0.92-1.03) ↔︎
Experimental Treatment (any) 1.96 (1.03-3.73) ⬆
Hyperkaliemia 1.83 (0.57-5.96) ↔︎
Hypokaliemia 6.49 (2.19-19.19) ⬆
Rhabdomyolysis 6.81 (3.44-13.47) ⬆
Severe Hypoglycemia 1.47 (0.41-5.22) ↔︎
Severe Thrombocytopenia 10.44 (1.72-63.2) ⬆
Abdominal pain 1.51 (0.83-2.74) ↔︎
Agitation 2.57 (0.89-7.43) ↔︎
Diarrhoea 1.02 (0.45-2.35) ↔︎
Hiccups 1.71 (0.52-5.62) ↔︎
Myalgia / Body Pain 1.43 (0.76-2.67) ↔︎
Nausea/Vomiting 1.11 (0.43-2.85) ↔︎
Signs of dehydration 0.92 (0.38-2.25) ↔︎

Logistic Regression of Factors Associated with Death and Complications

Final LASSO-Selected Logistic Regression: EVD Outcome
Multivariable model including LASSO-selected predictors
Characteristic N Event N OR 95% CI p-value Significance1 Direction Unadjusted RR2 Adjusted Risk Ratio (95% CI)3
CT at admission 354 134 0.73 0.67, 0.78 <0.001 *** ⬇
0.86 (0.84-0.89)
Delay from Onset to Admission 354 134 1.14 1.05, 1.23 0.001 ** ⬆
1.02 (1-1.04)
Hyperkaliemia 354 134 17.7 0.75, 1,120 0.17 ⬆ 2.32 1.27 (0.82-1.97)
Rhabdomyolysis 354 134 1.88 0.56, 6.77 0.31 ⬆ 1.89 1.16 (0.8-1.67)
Anuria anytime 354 134 2.64 0.53, 16.1 0.26 ⬆ 2.05 1.36 (0.8-2.32)
Difficulty breathing anytime 354 134 4.31 1.74, 11.4 0.002 ** ⬆ 1.91 1.58 (1.2-2.08)
Signs of dehydration anytime 354 134 0.29 0.11, 0.71 0.008 ** ⬇ 1.14 0.65 (0.47-0.91)
Unusual bleeding bruising anytime 354 134 2.36 1.19, 4.74 0.014 * ⬆ 1.85 1.32 (1.07-1.64)
Anorexia 354 134 0.62 0.28, 1.33 0.22 ⬇ 1.41 0.88 (0.65-1.19)
Diarrhoea 354 134 1.14 0.54, 2.38 0.74 ⬆ 1.79 1.14 (0.85-1.52)
Fatigue/weakness anytime 354 134 2.48 0.83, 7.95 0.11 ⬆ 2.71 1.51 (0.89-2.58)
Nausea/vomiting anytime 354 134 0.52 0.24, 1.13 0.10 . ⬇ 1.42 0.85 (0.6-1.19)
Sore throat/swallowing anytime 354 134 1.46 0.70, 3.05 0.32 ⬆ 1.52 1.06 (0.82-1.37)
Abbreviations: CI = Confidence Interval, OR = Odds Ratio, NA
1 Odds ratios (OR) obtained by exponentiating the logistic regression coefficients (β). OR > 1 indicates increased odds of the outcome; OR < 1 indicates decreased odds.
2 Raw risk ratios (RR) were calculated as the ratio of the risk of death among participants with the condition (Yes) versus without the condition (No).
3 Adjusted risk ratios (RR) obtained from Poisson regression with robust standard errors. RR > 1 indicates increased risk; RR < 1 indicates decreased risk.
  • For each 1 Ct increase at admission, the odds of death decreased by 27% (OR 0.73, 95% CI 0.67–0.78) and the risk of death decreased by 14% (RR 0.86).
  • For each day of delay at admission, the odds of death increased by 14% (OR 1.14, 95% CI 1.05–1.23) and the risk of death increases by 2% (RR 1.02).
  • Note: While the unadjusted risk of death was slightly higher in participants with dehydration, the multivariable model adjusting for other clinical factors indicated that dehydration was associated with lower risk of death (adjusted RR = 0.6), likely reflecting confounding factors.
Final LASSO-Selected Logistic Regression: Renal Impairment
Multivariable model including LASSO-selected predictors
Characteristic N Event N OR 95% CI p-value Significance1 Direction Unadjusted RR2 Adjusted Risk Ratio (95% CI)3
CT at admission 354 40 0.97 0.89, 1.05 0.45 ⬇
0.97 (0.92-1.03)
Experimental treatment 354 40 2.40 1.01, 5.91 0.051 . ⬆ 2.40 1.96 (1.03-3.73)
Hyperkaliemia 354 40 31.9 3.17, 355 0.003 ** ⬆ 7.08 1.83 (0.57-5.96)
Hypokaliemia 354 40 16.3 2.95, 84.4 <0.001 *** ⬆ 3.82 6.49 (2.19-19.19)
Rhabdomyolysis 354 40 38.8 11.5, 151 <0.001 *** ⬆ 9.49 6.81 (3.44-13.47)
Severe Hypoglycemia 354 40 10.3 0.64, 196 0.12 ⬆ 6.44 1.47 (0.41-5.22)
Severe thrombocytopenia 354 40 24.7 1.02, 318 0.016 * ⬆ 3.00 10.44 (1.72-63.2)
Adbominal pain 354 40 1.77 0.70, 4.53 0.23 ⬆ 1.88 1.51 (0.83-2.74)
Agitation 354 40 3.49 0.51, 18.1 0.16 ⬆ 2.53 2.57 (0.89-7.43)
Diarrhoea 354 40 1.41 0.46, 4.89 0.56 ⬆ 2.44 1.02 (0.45-2.35)
Hiccups 354 40 3.06 0.69, 10.9 0.10 ⬆ 1.80 1.71 (0.52-5.62)
Myalgia/Body pain 354 40 1.76 0.73, 4.47 0.22 ⬆ 2.04 1.43 (0.76-2.67)
Nausea/vomiting anytime 354 40 0.94 0.32, 3.08 0.92 ⬇ 2.42 1.11 (0.43-2.85)
Signs of dehydration anytime 354 40 1.25 0.35, 4.02 0.71 ⬆ 1.96 0.92 (0.38-2.25)
Abbreviations: CI = Confidence Interval, OR = Odds Ratio, NA
1 Odds ratios (OR) obtained by exponentiating the logistic regression coefficients (β). OR > 1 indicates increased odds of the outcome; OR < 1 indicates decreased odds.
2 Raw risk ratios (RR) were calculated as the ratio of the risk of death among participants with the condition (Yes) versus without the condition (No).
3 Adjusted risk ratios (RR) obtained from Poisson regression with robust standard errors. RR > 1 indicates increased risk; RR < 1 indicates decreased risk.
  • Patients with renal impairment had substantially higher odds of developing hyperkalemia compared to those with normal renal function. However, the confidence interval is wide, reflecting the small number of hyperkalemia events.
Final LASSO-Selected Logistic Regression: Bleeding
Multivariable model including LASSO-selected predictors
Characteristic N Event N OR 95% CI p-value Significance1 Direction Unadjusted RR2 Adjusted Risk Ratio (95% CI)3
Signs of dehydration anytime 354 36 3.06 1.27, 6.97 0.009 ** ⬆ 3.02 2.54 (1.27-5.07)
CT at admission 354 36 0.95 0.89, 1.01 0.088 . ⬇
0.95 (0.89-1.02)
Abbreviations: CI = Confidence Interval, OR = Odds Ratio, NA
1 Odds ratios (OR) obtained by exponentiating the logistic regression coefficients (β). OR > 1 indicates increased odds of the outcome; OR < 1 indicates decreased odds.
2 Raw risk ratios (RR) were calculated as the ratio of the risk of death among participants with the condition (Yes) versus without the condition (No).
3 Adjusted risk ratios (RR) obtained from Poisson regression with robust standard errors. RR > 1 indicates increased risk; RR < 1 indicates decreased risk.

Survival Analyses

  • We used a Weibull proportional hazards model to estimate survival probabilities over time.

  • Forest plot showing relative effects of covariates on survival time estimated from a Weibull proportional hazards model. HR > 1 indicates shorter survival (faster occurrence of the event), HR < 1 indicates longer survival. Error bars represent 95% confidence intervals.

Thank you